1. Field of the Invention
The present invention relates to an imaging apparatus for capturing a subject image.
2. Description of the Related Art
In recent years, digital still cameras that are mainly used for capturing still images and imaging devices having a maximum number of 10,000,000 pixels have come into use, and even for movie cameras, whose main application is to capture moving images, imaging devices of several million pixels have come into use.
It is clear that such increases in the number of pixels of imaging devices is attributable to the shrinking of the pixel size. However, as a result, the aperture area of the pixel also becomes smaller, and a low sensitivity and light shot noise become noticeable problems. Furthermore, as the number of electrons involved decreases, noise resulting from the imaging device become conspicuous.
As a method of reducing noise of an amplifying imaging device, it is common practice that, in order to remove reset noise at the input stage of an amplifying transistor, which appears as fixed pattern noise on the image capturing-plane and Vth variations of a transistor forming the amplifier, the noise is read in advance, and thereafter, a photoelectric conversion signal is superposed on the noise and read, and noise processing is performed by subtracting them, that is, so-called (S−N) processing is performed. In this method, it is recognized as common knowledge that fixed pattern noise can be reduced, but random noise becomes √2 times as great.
A conventional example in which such random noise is reduced is described below.
(1) As disclosed in U.S. Pat. No. 5,943,094, in the pixel noise reduction process, based on capturing output obtained in a light shielding state, predetermined noise data is generated from a plurality of pieces of noise data. For the captured data, data at one time after AD conversion after an image-capturing operation is performed is used. Since the predetermined noise data is averaged, the random noise is reduced. However, the random noise contained in the image-captured data cannot be reduced. Furthermore, some time for reading noise from the imaging device is required, the image-capturing interval becomes long, and image capturing cannot be performed immediately. Furthermore, the reactive power consumption becomes larger.
(2) As disclosed in U.S. Patent Publication 20020191742, in order to reduce X-ray noise, the photo-converted signal is read a plurality of times, and based on the difference, the X-ray noise is separated. In this case, random X-ray noise can be separated, but the pixel noise increases.
(3) As disclosed in U.S. Pat. No. 6,037,577, a method of reducing the influence of noise generated by a signal processing circuit at the subsequent stage by reading noise or a signal from an amplifying transistor a plurality of times and by increasing the noise and the signal amplitude by performing charge addition has been proposed. In, for example, FIG. 2 of the embodiment, charge addition is difficult in the case where, merely, a signal is electrically supplied to the capacitor after passing through a conventional MOS switch. When a charge-transfer MOS transistor is used, since the transfer of the signal requires a sufficient time, high-speed driving and high-speed continuous image capturing are difficult, and the possibility that 1/f noise of the amplifying transistor cannot be reduced is high.
(4) As disclosed in Japanese Patent Laid Open No. 2001-36920, a method of reducing noise in a reading-system signal processing circuit is known. In this case, in a method of reducing noise from the amplifying transistor, noise of the signal processing circuit at the subsequent stage can be ignored by a CDS (correlated double sampler) circuit and amplifier, which is effective in reducing the fixed pattern noise of the amplifying transistor.
In the foregoing, the conventional example is described with a view to reducing noise of the imaging device. In the applications of an imaging device having a large number of pixels, it is common practice that image-capturing is performed with a large number of pixels in high-precision image-capturing requiring a high resolution, and image-capturing is performed with a small number of pixels in a case in which a low resolution is sufficient. At that time, in the high-precision image-capturing, almost all of the pixel signals are read from the imaging device. In the low-resolution image capturing, for preventing battery consumption of the camera or for capturing moving images, pixel signals are read while being thinned out, or pixel signals are read while being thinned out and added.
In a first example of the above-described technology, as disclosed in U.S. Pat. No. 6,124,888, reading addition is performed by thinning out the pixels of the same color in units of 4×4 pixels. In a second example, disclosed in Japanese Patent Laid-Open No.2001-36920, by using 4×4 pixels as one group, a plurality of pixel signals are added so that the spatial color arrangement of each color before addition and the spatial color arrangement of each color after addition become the same.
In the publicly known example of noise reduction of the conventional amplifying imaging device, the fixed pattern noise caused by the amplifying transistor can be reduced to an image level at which no problem is posed in the (S−N) processing. As a result, 1/f noise (including random noise), which occurs when the amplifying transistor is driven, has become a problem. Or, it may be said that, finally, a technology level has been reached where the 1/f noise of the amplifying transistor after the S−N processing becomes a problem. This noise of the amplifying imaging device is less than or equal to that of a CCD in the high-precision image-capturing mode. However, in the addition reading of signals inside the imaging device, it has become a new problem. This is due to the following reasons. Since pixel addition in the CCD is charge addition, the signal is amplified as a result of the addition of the photoelectric conversion signals from a plurality of pixels. However, since the noise is determined by the amplifier (floating diffusion amplifier) at the final stage of the CCD, this case is the same as addition. However, in the amplifying imaging device, since a signal containing the noise for each pixel is added, noise becomes larger. It may be said that, when compared to the CCD, the amplifying imaging device has a poor SN ratio during dark time by √n times as large as the number of addition pixels. The 1/f noise of the amplifying transistor remains a significant problem.
Next, the problems of the method of adding a plurality of pixel signals are described.
In the above-described first example, the problem is that the number of effective pixels used from among the 4×4 pixels is small. In recent imaging devices, as a result of the imaging device having a larger number of pixels, the pixel unit size becomes smaller, and insufficient sensitivity has become a more significant problem. In the digital still camera, when a dark subject is captured, insufficient sensitivity can be compensated for by strobe light-emission, but during moving-image capturing, an expensive and heavy light source cannot be used, and significant noise occurs. As a result of thinning-out of the pixel signals, moire fringing due to a decrease in the sampling frequency occurs, and the deterioration of the image quality is severe.
In the above-described second example, the number of additions of the pixel signals within one group is increased, and sensitivity is improved. However, the problem is that pixel signals that are not used (discarded pixel signals) exist. A plurality of pixel signals are added so that the spatial color arrangement of each color becomes the same before and after addition within one group. However, another problem is that moire fringing occurs when the captured image is expanded.
As described above, in the conventional technology, since the pixel signals are thinned out, the sensitivity is not improved sufficiently, and moire fringing occurs even though the spatial color arrangement is identical.